This article examines strategy choices for how people find faults in a simple device by using models of several strategies and new data. Diag, a model solving this task, used a single strategy that predicted the behavior of most participants in a previous study with remarkable accuracy. This article explores additional strategies used in this reasoning task that arise when less directive instructions are provided. Based on our observations, five new strategies for the task were identified and described by being modeled. These different strategies, realized in different models, predict the speed of solution while the participant is learning the task, and were validated by comparing their predictions to the observations (r2 =.27 to.90). The results suggest that participants not only created different strategies for this simple fault-finding task but that some also, with practice, shifted between strategies. This research provides insights into how strategies are an important aspect of the variability in learning, illustrates the transfer of learning on a problem-by-problem level, and shows that the noisiness that most learning curves show can arise from differential transfer between problems.
All Science Journal Classification (ASJC) codes
- Experimental and Cognitive Psychology
- Cognitive Neuroscience
- Artificial Intelligence